A novel boundary extension approach for empirical mode decomposition

Abstract

The Empirical Mode Decomposition (EMD) proposed by Huang et al. in 1998 shows remarkably effective in analyzing nonlinear signals. However, the boundary extension is one of the theoretical problems unsolved in EMD. In this paper, a novel boundary processing technique is proposed to deal with the border effect in EMD. An algorithm based on the sigma-pi neural network is used to extend signals before applying EMD. By virtue of this method, the frequency compression near the end is eliminated and errors caused by end effect are reduced. Verifications of the experimental signals show that the newly proposed boundary extension method is useful in practice. © Springer-Verlag Berlin Heidelberg 2006.